Hi,

I would like to know which technique is more acceptable and speed for : 10 000 000 observations * 12 period * 500 kpi (the goal is to have 10 000 000 observations * 6000 kpi)

technique 1 :

data foo; input id1 id2 id3 X Y Z; cards; 1 1 1 43 54 665 1 1 2 45 6565 6565 run;

proc transpose data=foo out=foo2; by id1 id2 id3; var x y z; run;

data foo3; set foo2; myname = trim(_name_) || put(id3,4.-L); run;

proc transpose data=foo3 out=foowide; by id1 id2; id myname; var col1; run;

technique 2 :

data a; do ident=1 to 10; do periode=1 to 12; kpi1 = round(ranuni(123)*1000); kpi2 = round(ranuni(123)*1000); kpi3 = round(ranuni(123)*1000); kpi4 = round(ranuni(123)*1000); output; end; end; run;

data final; set a; by ident; retain final1-final48 ; array kpi(4); array final(48) ; if first.ident then do; do i=1 to 48; final(i)=.; end; end; do i=1 to dim(kpi); final(i+4*(periode-1))=kpi(i); end; if last.ident then output; run;

Thanks in advance, Philippe